Heart Rate During the Mock Audition
performance-anxiety
heart-rate
descriptive
MFA-2026
This analysis presents the heart rate data from five mock audition sessions. Each singer performed two songs – one prepared, one assigned 24 hours in advance – for a three-person faculty panel. All subjects wore Polar H10 heart rate monitors. HR was recorded at 1 Hz (one value per second) and aligned to audio using a clap-based sync point.
Participants:
- Singer 1 (S1, M)
- Singer 2 (S2, F)
- Singer 3 (S3, F)
- Singer 4 (S4, F)
- Singer 5 (S5, M)
- Panelist 1 (F1) – HR data available for all 5 sessions
- Panelist 2 (F2) – HR data available for all 5 sessions
- Panelist 3 (F3) – HR data available for all 5 sessions
Assigned pieces: Female singers (S2, S3, S4): “If You Knew My Story” | Male singers (S1, S5): “Shiksa Goddess”
Data constraint: HR was exported from Polar Flow at 1 Hz – averaged BPM per second, not beat-to-beat RR intervals. Standard HRV metrics cannot be computed. Analysis focuses on HR magnitude, reactivity, and interpersonal synchrony.
Methodological notes: - All faculty panelists were seated throughout all auditions. No standing, walking, or postural changes occurred. All panelist HR variations reflect genuine autonomic responses, not movement artifacts. - Panelist 1 wore a facial mask for the duration of the experiment to avoid sharing or receiving illness. This may have contributed slightly to Panelist 1’s elevated baseline HR (~100 bpm), though the consistency across all five sessions suggests a stable individual difference rather than a masking artifact.
Part 1: Individual Session Time Series
Each figure shows one audition session with the singer’s HR (red) overlaid with Panelist 1 (purple), Panelist 2 (green), and Panelist 3 (blue). Shaded bands indicate audition phases.
Panelist baseline context: The three panelists – all seated throughout – show a stable hierarchy across all sessions: Panelist 1 runs highest (~98-108 bpm), Panelist 3 is moderate (~75-95 bpm), and Panelist 2 is lowest (~55-65 bpm). Panelist 2’s HR is essentially resting heart rate for a seated adult. These individual differences in resting autonomic tone are consistent across all five sessions, suggesting stable trait-level variation rather than differential responses to specific singers.
Session 1 – Singer 1 (M)
Singer 1 begins around 127 bpm and ramps steadily upward through song 1 (prepared piece), reaching peaks near 160 bpm. There is a brief dip during the inter-song gap, followed by a sharp rise into song 2 (assigned piece), where HR reaches its highest sustained values (~155-158 bpm). The three panelists hold flat: Panelist 1 around 100-108 bpm, Panelist 3 at 85-95 bpm, and Panelist 2 near 58 bpm. The singer’s activation gap above the panelist mean is ~60-70 bpm during performance.
Session 2 – Singer 2 (F)
Singer 2 shows the steepest ramp-up of any singer – HR climbs from ~120 bpm at the start to peak near 169 bpm during song 1 (prepared). HR drops modestly during the inter-song gap (~145 bpm) but rebounds during song 2 (assigned), reaching ~155 bpm. Panelist 1 drifts from 108 down to ~100 bpm; Panelist 3 hovers in the 75-90 bpm range; Panelist 2 holds steady near 57 bpm. The activation gap widens as the session progresses.
Session 3 – Singer 3 (F)
Singer 3 follows a similar arc to Singer 2 – gradual climb through the pre-singing phase, sustained elevation during song 1 (135-165 bpm), partial recovery in the gap, and a plateau during song 2 (~150-158 bpm). Panelist 1 is steady around 101-107 bpm; Panelist 3 is at 80-95 bpm; Panelist 2 stays near 58-60 bpm. The inter-song gap shows a brief but noticeable HR spike in Panelist 3’s trace – likely anticipatory activation as the panelist prepares for the next performance (see Panelist Anticipatory Activation below).
Session 4 – Singer 4 (F)
Singer 4 is a clear outlier. HR starts around 80-90 bpm in the pre-singing phase – within the range typically seen in a panelist, not a singer. During song 1, HR rises modestly to ~105-120 bpm, well below the 140-170 bpm range seen in other singers. During song 2, HR actually drops below baseline, reaching the mid-70s. This is the most visually striking session in the dataset: the singer’s trace overlaps with Panelist 1 and Panelist 3, while Panelist 2 runs below all of them. Panelist 1 spikes to ~113 bpm during the inter-song gap – the highest F1 variability in any session (sd=5.8). Panelist 3 is similarly variable (sd=8.1 during the gap).
Important context: Panelist 3 had a pre-existing private studio teaching relationship with Singer 4. Panelist 1 had previously taught all five participants in a classroom setting. Singer 4’s reduced HR may reflect the familiarity of the studio teacher-student relationship, individual differences in autonomic reactivity, anxiolytic medication use (not screened for), or some combination.
Session 5 – Singer 5 (M)
Singer 5 shows the highest sustained HR of any participant. Starting from a pre-singing baseline of ~135-140 bpm (already elevated), HR climbs rapidly through song 1 to peaks near 184 bpm. The inter-song gap provides minimal recovery (~165 bpm), and song 2 drives HR to its highest sustained values (~170-180 bpm). The three panelists form a clear floor: Panelist 1 at 96-106 bpm, Panelist 3 at 73-85 bpm, Panelist 2 at 54-59 bpm. The widest activation gap in the study occurs here – ~91 bpm between singer and the 3-panelist mean during song 2.
Part 2: Cross-Session Comparisons
Singer HR by Audition Phase
The grouped bar chart below shows each singer’s mean HR (with standard deviation bars) during the four performance phases. The pattern is consistent across four of five singers: HR is lowest during pre-singing, rises sharply for song 1, dips slightly during the inter-song gap, and remains elevated or continues rising into song 2. Singer 4 is the exception – her HR stays in the 88-106 bpm range across all phases, never approaching the 140+ bpm levels seen in the others.
Key observation: Pre-singing baseline HR is remarkably similar across Singers 1, 2, 3, and 5 (~126-143 bpm), suggesting comparable levels of anticipatory anxiety before singing begins. Singer 4’s pre-singing HR (~91 bpm) is 35-50 bpm lower than the group.
Singer-Panelist Activation Gap
The chart below quantifies the HR difference between each singer and the mean of all three panelists for each phase. The gap reflects how far the singer’s autonomic activation exceeds the panelists’ – a rough index of the anxiety asymmetry in the audition relationship. With Panelist 2 included (mean HR ~58 bpm, seated), the 3-panelist average is lower than the earlier 2-panelist estimate, so activation gaps are wider.
Singers 1, 2, 3, and 5 show gaps of 38-91 bpm, widening as the session progresses. The gap is largest during the inter-song gap and song 2 for most singers – suggesting that the stress of performing the assigned piece (less familiar, less rehearsed) compounds the already elevated state from song 1.
Singer 4’s gap is small but positive (+8 bpm during song 2) – her HR stays close to the panelist range. This is the only singer whose physiological state approximates the panelists’.
HR Reactivity Profiles
The dot-and-line plot below connects each singer’s mean HR across the four phases, making the shape of each stress response visible.
Four distinct profiles emerge:
- Singer 5 – highest absolute HR, steep climb, no recovery between songs. The most pronounced stress response in the study.
- Singer 2 – steep initial climb during song 1, partial recovery in the gap, rebound for song 2. Classic performance anxiety arc.
- Singers 1 and 3 – moderate, steady climb. Singer 1 shows the most gradual escalation; Singer 3 plateaus earlier.
- Singer 4 – flat trajectory near 90 bpm. Modest rise during song 1, return to baseline for song 2. This profile resembles a panelist more than a performer.
Normalized HR Trajectories
All five singers’ HR traces plotted on a common timeline, each zeroed to their own pre-singing baseline mean. This removes differences in resting HR and shows the magnitude of change from baseline.
With a 10-second rolling average for readability:
- Singers 2 and 5 show the largest sustained elevations (~25-40 bpm above baseline).
- Singer 1 shows a moderate, steady rise (~15-30 bpm above baseline).
- Singer 3 shows a moderate rise with more variability.
- Singer 4 shows the smallest reactivity – peaking around +15 bpm during song 1, then dropping below baseline during song 2.
Part 3: Self-Report Measures: K-MPAI, MAAQ, and Post-Surveys
This section presents the self-report data that frame the physiological findings: how anxious these singers report being (trait measures), how they experienced the audition (singer post-surveys), and how the panelists perceived them (faculty post-surveys). The correlations between these measures and HR are presented in Part 4.
K-MPAI and MAAQ Pre-Survey Results
The K-MPAI was administered to all BoCo BFA/MM voice students as a pre-screening survey. The 5 study participants were drawn from this broader pool of 26 respondents. Scoring follows Kenny (2009): 40 items on a 0-6 scale with 8 reverse-coded positive items (1, 2, 9, 17, 23, 33, 35, 37), range 0-240, clinical threshold >=105.
Population (26 BoCo singer respondents):
- Mean K-MPAI: 124.1 (SD 31.6), Median: 124.0, Range: 64-185
- 73% (19/26) score >=105 (clinically significant MPA)
- This is a high-anxiety population, consistent with conservatory performance culture
Study participants:
| Singer | K-MPAI | Clinical? | MAAQ Flexibility | Notes |
|---|---|---|---|---|
| S1 (M) | 126 | Yes | 31 | Near population mean |
| S2 (F) | 111 | Yes | 35 (highest) | Lowest anxiety, highest flexibility |
| S3 (F) | 144 | Yes | 26 | Above population mean |
| S4 (F) | 159 | Yes | 21 | High anxiety despite lowest HR |
| S5 (M) | 174 | Yes | 17 (lowest) | Highest anxiety, lowest flexibility |
All 5 singers score above the clinical threshold – not surprising given they self-selected for a study on performance anxiety. The rank ordering is notable: Singer 5 (highest K-MPAI, 174) also shows the highest sustained HR in the study, while Singer 2 (lowest K-MPAI, 111) shows the steepest HR ramp-up but also the most recovery between songs. Singer 4 (K-MPAI 159 – second highest) is the most interesting case: high self-reported anxiety with the lowest physiological response. Several factors may contribute, including the studio relationship with Panelist 3, individual autonomic differences, or anxiolytic medication (not screened for in this study).
Faculty panelists:
| Panelist | K-MPAI | Clinical? | MAAQ Flexibility | Mean HR (seated) |
|---|---|---|---|---|
| Panelist 1 | 109 (2 items missing) | Yes | 24 | ~101 bpm |
| Panelist 2 | 102 | No | 33 | ~58 bpm |
| Panelist 3 | 169 | Yes | 32 | ~85 bpm |
Panelist 3’s K-MPAI (169) is notably high – higher than any study participant except Singer 5. For a panelist whose own performance anxiety is elevated, evaluating singers (especially her own student, Singer 4) may carry additional physiological weight. Panelist 2 falls just below the clinical threshold at 102 and shows the lowest HR of anyone in the study – essentially resting heart rate while seated. The K-MPAI ranking on the faculty side (F3 > F1 > F2) does not map directly to their HR ranking (F1 > F3 > F2), but Panelist 2 being the calmest on both measures is consistent.
Limitation: The study did not screen for anxiolytic medication use (beta blockers, SSRIs, benzodiazepines) in singers or faculty panelists. Any participant using such medication could show attenuated HR responses that do not reflect their underlying anxiety level.
K-MPAI Scoring Note
The Qualtrics-generated SC0 score differs from our computed score by a constant of +40 across all 26 respondents (verified: identical SDs, perfect rank correlation). Qualtrics scores the K-MPAI on a 1-7 internal scale, while Kenny (2009) specifies 0-6. Both apply the same reverse coding on the same 8 items – the only difference is a +1 shift per item across all 40 items (40 x 1 = 40). We use the published 0-6 convention. The Qualtrics threshold equivalent of >=105 is >=145.
Part 4: Anxiety Measures and Heart Rate
This section connects the self-report data from Part 3 to the physiological record.
K-MPAI x Heart Rate (RQ2)
Does higher trait performance anxiety predict higher physiological arousal during the audition? With n = 5 singers, Spearman rank correlations are used. The critical value for significance at p < 0.05 (two-tailed) with n = 5 is |r_s| = 1.000, so no correlation can reach conventional significance. These are descriptive rank-order patterns, not inferential tests.
| Anxiety Measure | HR Measure | r_s | p |
|---|---|---|---|
| K-MPAI Total | Mean Song HR | +0.10 | 0.873 |
| K-MPAI Total | Peak HR | +0.10 | 0.873 |
| K-MPAI Total | HR Reactivity | -0.30 | 0.624 |
| K-MPAI Total | Activation Gap | +0.10 | 0.873 |
| MAAQ Flexibility | Mean Song HR | -0.10 | 0.873 |
| MAAQ Raw Sum | Mean HR (Assigned Song) | -0.87 | 0.054 |
No meaningful K-MPAI-HR association. Singer 4’s position breaks any monotonic pattern: she has the second-highest K-MPAI (159) but the lowest HR across every measure. If Singer 4 were removed, the remaining four singers show a near-monotonic positive relationship between K-MPAI and Mean Song HR. Several factors may contribute to Singer 4’s low HR: the pre-existing studio teaching relationship with Panelist 3, individual differences in autonomic reactivity, or anxiolytic medication use (not screened for in this study). With n = 5, a single case can shift the entire correlation.
Post-Survey x Heart Rate Correlations
Singer Self-Report
Singers completed an 8-item post-performance survey (1-5 Likert) immediately after their audition. Each item is correlated with their mean song HR.
Critical finding: “Felt anxious” has zero variance. All five singers rated their anxiety at 4 out of 5. With no rank-order variance, the correlation is undefined. This is itself a substantive result: every singer, including Singer 4 (whose HR was near panelist levels), self-reported high anxiety. The audition context produced a universal subjective experience of anxiety regardless of the physiological response, consistent with research showing that cognitive and somatic anxiety components can dissociate.
| Survey Item | Direction | vs Mean Song HR (r_s) |
|---|---|---|
| Felt confident | positive | +0.35 |
| Felt anxious | negative | NaN (no variance) |
| Felt authentic | positive | 0.00 |
| Felt safe to express | positive | 0.00 |
| Emotionally connected | positive | +0.78 |
| Emotionally disconnected | negative | -0.05 |
| Enjoyed performing | positive | +0.37 |
| Did not feel satisfaction | negative | +0.63 |
“Emotionally connected” shows the strongest positive correlation with HR (r_s = +0.78, p = 0.118). Singers with higher HR reported feeling more emotionally connected to their performance, not less. This is consistent with the idea that physiological activation during performance is not purely aversive: it can coexist with engagement and emotional presence.
Faculty Perception x Singer HR
Faculty panelists rated each singer on 6 items after each audition. All three panelists watched each performance simultaneously, so ratings of a single singer’s performance are correlated.
Primary Analysis: Singer-Averaged Ratings (n = 5)
The 3 panelists’ ratings are averaged per singer, then correlated with singer HR. This gives n = 5 independent observations.
| Faculty Item | vs Singer Mean HR (r_s) | p |
|---|---|---|
| Appeared confident | +0.50 | 0.391 |
| Showed signs of anxiety | -0.70 | 0.188 |
| Felt authentic | -0.31 | 0.614 |
| Emotionally connected | -0.26 | 0.668 |
| Emotionally disconnected | -0.10 | 0.870 |
| Appeared to enjoy performing | +0.58 | 0.306 |
Key observations:
“Appeared confident” x Singer HR retains a positive direction (r_s = +0.50) but with an honest p = 0.391. Panelists rated higher-HR singers as appearing more confident, not less. The n=15 pooled analysis (below) showed p = 0.057, but that p-value is anti-conservative due to non-independence.
“Showed signs of anxiety” x Singer HR is the strongest faculty-perception correlation (r_s = -0.70). Higher-HR singers were rated as showing fewer signs of anxiety. Physiological stress was not visible to the panelists.
Per-panelist breakdown reveals that Panelist 2 drives the “Appeared confident” pattern (r_s = +0.87), while Panelist 3 shows no relationship (r_s = -0.11). Panelist 3 (highest K-MPAI, Singer 4’s studio teacher) does not perceive high-HR singers as more confident. Panelist 2 (lowest anxiety, most stable HR) shows the strongest perceptual tracking.
Supplementary: Pooled (n = 15, non-independent, anti-conservative p-values)
| Faculty Item | vs Singer Mean HR (r_s) | vs Panelist Own HR (r_s) |
|---|---|---|
| Appeared confident | +0.50 (p=.057) | -0.14 |
| Showed signs of anxiety | -0.40 (p=.138) | -0.12 |
| Felt authentic | -0.03 | -0.20 |
| Emotionally connected | -0.07 | -0.03 |
| Emotionally disconnected | -0.09 | +0.12 |
| Appeared to enjoy performing | +0.34 | +0.02 |
Panelist own HR shows no meaningful correlation with any perception item.
Trajectory Features
The mean-level correlations above collapse each singer’s entire song performance into a single HR value. This section preserves temporal structure by extracting trajectory features from HR time series during each song phase and fitting individual growth curves.
Trajectory Features
Five features were extracted from each singer’s HR during song_1 and song_2, then averaged across both songs:
- Slope: linear regression slope of HR over normalized time (0 = song onset, 1 = song offset). Using normalized time avoids a confound with song duration.
- AUC above baseline: area under the HR curve above the singer’s pre-singing baseline HR, normalized by duration (bpm above resting).
- Time to Peak: seconds from phase start to maximum HR
- HR Range: max minus min HR within the phase (amplitude of fluctuation)
- Recovery: HR at phase end minus HR at phase start (positive = continued climbing, negative = recovery)
| Feature | K-MPAI r_s | p | MAAQ Flex. r_s | p |
|---|---|---|---|---|
| Slope (bpm/norm. time) | +0.70 | 0.188 | -0.70 | 0.188 |
| AUC above baseline (bpm) | -0.30 | 0.624 | +0.30 | 0.624 |
| Time to Peak | +0.30 | 0.624 | -0.30 | 0.624 |
| HR Range | +0.90 | 0.037 | -0.90 | 0.037 |
| Recovery | +0.70 | 0.188 | -0.70 | 0.188 |
Key findings:
- HR Range is the strongest trajectory-anxiety association (r_s = +0.90, p = 0.037, uncorrected). Higher trait anxiety predicts wider HR fluctuation during singing. Slope and Recovery also show positive associations with K-MPAI (r_s = +0.70 each) but do not reach conventional significance. These trajectory features reveal information that mean HR obscures. Unlike the near-zero K-MPAI x Mean Song HR correlation (r_s = +0.10), the trajectory features capture the dynamic quality of the stress response.
Note: All p-values are uncorrected for multiple comparisons. Across the full analysis, over 50 correlations are computed. Even the one p = 0.037 result (HR Range) would not survive Bonferroni correction. All correlations should be treated as exploratory pattern descriptions.
Singer 4 no longer breaks the pattern for HR Range. In the mean HR analyses, Singer 4’s low absolute HR (despite high K-MPAI) produced flat correlations. But Singer 4’s range is actually consistent with moderate-to-high K-MPAI: she shows meaningful HR fluctuation (23-40 bpm range) even though her absolute HR stays low. The trajectory features reveal that Singer 4’s autonomic system is responding to the performance, just from a lower set point.
Recovery is positive for all singers. No singer shows net HR recovery within either song phase. All singers’ HR is still climbing or elevated at the end of each song.
Part 5: Individual Growth Curves
For each singer and song phase, HR was modeled over normalized time (0 = song onset, 1 = song offset) using both linear and quadratic regression. The quadratic fit captures inverted-U or U-shaped trajectories that a linear model would miss.
| Singer | Song | Shape | Linear R2 | Quad R2 | Notes |
|---|---|---|---|---|---|
| S1 | Prepared | Ramp-up | 0.852 | 0.861 | Steady linear climb |
| S1 | Assigned | Ramp-up | 0.733 | 0.736 | Steady linear climb |
| S2 | Prepared | Inverted-U | 0.044 | 0.913 | Classic arc: rise then fall |
| S2 | Assigned | Ramp-up | 0.353 | 0.384 | Shifts to climb pattern for new piece |
| S3 | Prepared | Inverted-U | 0.739 | 0.794 | Rise with late plateau |
| S3 | Assigned | Inverted-U | 0.418 | 0.599 | Moderate arc |
| S4 | Prepared | Inverted-U | 0.013 | 0.428 | Rise and fall within panelist range |
| S4 | Assigned | Ramp-up | 0.889 | 0.889 | Pure linear climb from low baseline |
| S5 | Prepared | Inverted-U | 0.797 | 0.894 | Steep climb with late plateau near 180 bpm |
| S5 | Assigned | U-shape | 0.799 | 0.902 | Brief early dip then steep climb |
Interpretive notes:
S2 prepared piece is the most dramatic inverted-U (R2 improvement = +0.87). A linear model explains almost nothing (R2 = 0.04), while the quadratic captures 91% of variance.
The prepared-to-assigned shift changes trajectory shape. Three singers (S2, S4, S5) show different shapes between songs. The assigned piece may not allow enough time or familiarity for the downslope of the inverted-U to emerge.
S1 is the most consistently linear. Both songs show high linear R2 (0.73-0.85) with negligible quadratic improvement.
Part 6: Same-Song Comparisons
A. Same-Song Comparisons with Panelists
Because the assigned piece was the same within each gender group, we can compare singers’ HR during the identical musical material. Any differences in HR trajectory during the same song reflect individual anxiety responses rather than differences in musical demands.
Female Singers – "If You Knew My Story"
During the assigned piece (right panel) – the same 32-bar cut – Singers 2 and 3 track closely in the 140-155 bpm range with similar contour, while Singer 4 starts at ~80 bpm and rises slowly to ~99 bpm. The separation between Singer 4 and the other two is even more striking here than during the prepared pieces, because the musical demands are identical.
Each female singer during "If You Knew My Story" with all three panelists overlaid:
- Singer 2: Sustained at 140-155 bpm. All three panelists hold flat below, with Panelist 2 running lowest (~57 bpm). The singer sits 40-80+ bpm above the panelist range.
- Singer 3: Similar profile to Singer 2. Panelist traces are stable and well below.
- Singer 4: Starts at ~80 bpm – below Panelist 1 (~98 bpm) and near Panelist 3 (~92 bpm). Only Panelist 2 (~56 bpm) is consistently below her. This is the only session where the singer’s HR is intertwined with the panelists’ during active singing.
Male Singers – "Shiksa Goddess"
During the assigned piece (right panel), Singer 5 starts at ~166 bpm and climbs to ~180 bpm. Singer 1 starts at ~149 bpm and rises gently to ~158 bpm. The ~15-20 bpm gap between them is consistent, suggesting a stable difference in stress reactivity rather than a response to a specific musical moment.
Each male singer during "Shiksa Goddess" with all three panelists:
- Singer 1: Flat trace at ~150-158 bpm. All three panelists well below. The gap between singer and 3-panelist mean (~70 bpm) is stable throughout.
- Singer 5: Climbing trace from ~166 to ~180 bpm. The gap widens over the course of the song – panelists hold flat or decrease while Singer 5 continues climbing. Panelist 3 shows the clearest antiphase pattern (HR decreasing as singer’s increases).
B. Singer-to-Singer HR During Identical Pieces
Part 6A showed each singer alongside their panelists during the same song. This subsection removes the panelists and asks: how do singers’ HR time series compare when they perform the same song?
The assigned piece (song 2) provides a natural control. All three female singers (S2, S3, S4) sang “If You Knew My Story,” and both male singers (S1, S5) sang “Shiksa Goddess.” Because the song is identical within each group, the respiratory demands, musical structure, phrasing, and emotional arc are held constant. Any commonalities in HR trajectory shape reflect song-driven physiological effects. Any differences in absolute HR level reflect individual stress response.
Song 2 durations are tightly clustered (54-58 seconds). Both absolute-time alignment (from song onset) and normalized-time alignment (0-1) are used. Pairwise Pearson correlations are computed on the raw, normalized, and 5-second smoothed time series.
Male Singers: “Shiksa Goddess” (S1 and S5)
| Singer | Duration | HR Range | Mean HR | K-MPAI |
|---|---|---|---|---|
| S1 (M) | 58s | 149-158 | 154.4 | 126 |
| S5 (M) | 58s | 163-182 | 169.9 | 174 |
Pairwise correlation: r = +0.80 (normalized), r = +0.84 (smoothed).
S1 and S5 produce nearly parallel HR trajectories: a steady upward ramp with a slight dip around the 30-40% mark, then a sustained climb to the end. S5 is offset approximately 15 bpm above S1 throughout. The shape is almost identical; the offset tracks with K-MPAI (174 vs. 126). Between-singer SD on the mean trajectory is only 7.8 bpm.
Female Singers: “If You Knew My Story” (S2, S3, S4)
| Singer | Duration | HR Range | Mean HR | K-MPAI |
|---|---|---|---|---|
| S2 (F) | 56s | 134-157 | 147.4 | 111 |
| S3 (F) | 55s | 141-158 | 151.8 | 144 |
| S4 (F) | 54s | 76-99 | 89.5 | 159 |
Pairwise correlations (normalized time):
| Pair | r | r (smoothed) |
|---|---|---|
| S2-S3 | +0.27 | +0.29 |
| S2-S4 | +0.68 | +0.71 |
| S3-S4 | +0.70 | +0.73 |
S3 and S4 are strongly correlated (r = +0.70) despite being 60 bpm apart. S2-S4 similarly correlated (r = +0.68). But S2 and S3, despite similar absolute HR levels (147 vs. 152 bpm), correlate only weakly (r = +0.27), suggesting that at similar activation levels, individual-level factors (breath management, performance strategy) can override the song-driven pattern.
What This Shows
The song imposes a shared HR contour. Singers performing the same piece correlate at r = +0.68 to +0.84 (smoothed). The musical material drives a common autonomic pattern through its respiratory demands, phrasing structure, and intensity arc.
The individual stress response determines the set point, not the shape. S4 is 60 bpm below the other female singers but tracks the same contour as S3. S5 is 15 bpm above S1 but follows the same shape. K-MPAI predicts how high a singer’s HR is. The song determines when it rises and falls.
This helps explain the singer-panelist synchrony from Parts 8-10. The song drives a structured HR response in the singer. The panelists hear the same song. This likely accounts for the trend-level coherence concentrated at longer timescales: both singer and panelists respond to the musical arc, but through different mechanisms (respiratory demand vs. perceptual engagement).
Part 7: The S4-F3 Dyad and Panelist Patterns
The Singer 4 - Panelist 3 Relationship
Singer 4 is a student of Panelist 3 – they have a pre-existing private studio teacher-student relationship. Panelist 1 also had prior classroom teaching experience with all five participants, though in a group setting rather than one-on-one. These relational contexts are important interpretive lenses, but cannot be isolated as causal explanations for Singer 4’s low HR. Several factors may contribute:
Singer 4’s low HR is consistent with reduced evaluative threat. Performing for one’s own studio teacher – someone who knows your voice, has heard you at your worst, and is invested in your success – may reduce the sense that resources are insufficient to meet demands (the "threat" state in the biopsychosocial model; Guyon et al., 2020). However, the other singers also performed for a familiar classroom teacher (Panelist 1) without a comparable reduction, suggesting that the studio relationship’s intimacy may matter, or that individual differences in autonomic reactivity or coping style contribute independently.
The HR convergence between Singer 4 and Panelist 3 is consistent with what Coutinho et al. (2021) describe in established relationships. While Singer 4 and Panelist 3 are not romantic partners, the private studio relationship represents the closest interpersonal bond of any singer-panelist dyad in this study, and it produces the most convergent physiological profile.
All three panelists show elevated variability during Singer 4’s session. Panelist 3’s HR is the most variable of any session (sd=8.1 during the inter-song gap). Panelist 1 also shows its highest variability here (sd=5.8, spiking to ~113 bpm). Even Panelist 2, typically the most stable, shows slightly more variation.
For the thesis: This dyad should be discussed as both a limitation (the pre-existing relationships are confounds) and as a case study in co-regulation (the strongest evidence of physiological coupling occurs in the dyad with the deepest relational history). The explanation should be framed as one plausible factor rather than a confirmed mechanism, given that individual differences in autonomic reactivity or anxiolytic medication use (not screened for) could also account for Singer 4’s profile.
Anticipatory Anxiety
Four of five singers show pre-singing HR values of 126-143 bpm – substantially elevated above typical resting heart rate (60-80 bpm for young adults). This is consistent with Vellers et al. (2017), who identified the pre-performance anticipatory period as the most sensitive window for detecting audition stress. The singers are already in a state of sympathetic activation before they begin singing. Singer 4, again, is the exception at ~91 bpm.
Panelist Anticipatory Activation
The singer anticipatory anxiety finding has a counterpart on the panelist side. All three panelists – seated throughout – show HR ramp-ups in the final seconds of the inter-song gap before song 2 begins. Because the panelists did not stand, shift position, or move during the auditions, these HR changes cannot be attributed to postural artifacts. They reflect genuine autonomic activation – the evaluative experience of preparing to hear the next piece.
Panelist 3 shows the most consistent pattern (anticipatory activation in 4 of 5 sessions), with Session 4 (her own student) producing the largest ramp (+17.7 bpm). Panelist 1 shows activation in 3 of 5 sessions. Panelist 2, with the lowest overall HR, shows modest changes that are harder to distinguish from baseline fluctuation.
Prepared vs. Assigned Song
The data does not show a simple pattern of "assigned piece = more anxiety." Some singers (1, 3, 5) show higher HR during song 2 (assigned), but Singer 2 shows lower HR during song 2 than song 1. This may reflect:
- Fatigue or habituation (HR naturally declining after sustained high effort)
- The assigned piece being shorter and less demanding than some prepared pieces
- Individual differences in how novelty vs. familiarity affects stress
The comparison is also confounded by order – song 2 always follows song 1, so any "assigned piece" effect is inseparable from a "second performance" effect. This should be acknowledged as a design limitation.
Part 8: Physiological Synchrony (Windowed Cross-Correlation)
This section quantifies the degree to which singer and panelist HR move together during performance, using windowed cross-correlation (WCC). This addresses RQ1: does physiological coupling (synchrony) occur between singer and panelist during the audition?
Method
Following Moulder et al. (2018), WCC was computed with: - Window: 10 seconds (captures meaningful HR fluctuation at 1 Hz) - Step: 5 seconds (50% overlap) - Max lag: +/-8 seconds (physiological response latency) - Primary metric: mean detrended Pearson r at lag 0
Within each window, both signals are linearly detrended before correlation. Windows where either signal has SD < 0.5 bpm are excluded (the “near-constant guard,” primarily affecting Panelist 2).
Surrogate testing: Each singer’s HR was paired with each panelist’s HR from other sessions (session-swap). This preserves temporal structure while breaking real coupling.
Panelist 2 Exclusion Rate
Panelist 2’s HR was near-constant in 95% of windows (188 of 198 excluded). Panelist 2 cannot contribute meaningful synchrony data, confirming the near-resting physiological state observed in Parts 1-2.
WCC Results
| Singer | Panelist 1 (mean r) | Panelist 3 (mean r) | Panelist 2 (mean r) | Notes |
|---|---|---|---|---|
| S1 | +0.23 | +0.23 | +0.04 | Moderate positive synchrony |
| S2 | +0.19 | -0.08 | -0.36 | Positive with Panelist 1 only |
| S3 | +0.07 | -0.01 | +0.03 | Near zero across all panelists |
| S4 | +0.36 | +0.03 | +0.45 | Strongest Panelist 1 coupling |
| S5 | +0.03 | -0.15 | NaN | Panelist 3 shows antiphase pattern |
Key observations:
Singer 4 shows the strongest synchrony, but with Panelist 1, not Panelist 3. The S4-Panelist 1 pair has the highest mean r across both songs (+0.36), with particularly strong coupling during song_2 (r = +0.55). The studio teacher-student pair (S4-Panelist 3) shows near-zero coupling (+0.03).
Singer 5 and Panelist 3 show the only consistent antiphase pattern (mean r = -0.15). Panelist 3’s HR tends to move opposite to Singer 5’s during performance.
Most real pairs do not clearly exceed surrogates. With n = 5 and short song durations, statistical power is low. This question remains open for larger studies.
Synchrony and K-MPAI
| Comparison | r_s | p |
|---|---|---|
| K-MPAI vs Mean Sync (all panelists) | -0.10 | 0.873 |
| K-MPAI vs Sync with Panelist 1 | -0.30 | 0.624 |
| K-MPAI vs Sync with Panelist 3 | -0.30 | 0.624 |
| MAAQ Flex vs Mean Sync | +0.10 | 0.873 |
No meaningful relationship between trait anxiety and physiological synchrony. The direction is weakly negative (higher K-MPAI = slightly less synchrony), but the magnitudes are negligible.
Synchrony by Condition: Prepared vs. Assigned Song
The paired dot plot below shows each singer-panelist pair’s synchrony during Song 1 (prepared) vs. Song 2 (assigned). No consistent pattern emerges: some pairs show higher synchrony during the prepared piece, others during the assigned piece. The short duration of Song 2 (~55-59 seconds) limits the number of WCC windows and makes these estimates noisier.
Part 9: Trend-Level Cardiac Coherence (Smoothed Whole-Series Correlation)
The WCC analysis (Part 8) detrends each 10-second window before computing correlations, targeting moment-to-moment fluctuation coupling. However, visual inspection of the HR time series suggests a broader pattern: singer and panelist HR traces often share the same general contour over the course of a song, rising and falling together at a timescale of 15-60 seconds. The WCC detrending removes exactly this kind of slow co-movement.
To test whether this trend-level coherence is real, both HR signals are smoothed with a 15-second centered moving average, then a single Pearson correlation is computed on the full smoothed series for each singer-panelist-song pair.
| Pair | Song 1 r | Song 2 r | Mean r |
|---|---|---|---|
| S1-F1 | +0.391 | -0.838 | -0.224 |
| S1-F3 | +0.537 | -0.940 | -0.202 |
| S2-F1 | -0.860 | +0.857 | -0.002 |
| S2-F3 | +0.644 | +0.897 | +0.771 |
| S3-F1 | +0.462 | +0.857 | +0.660 |
| S3-F3 | +0.541 | +0.337 | +0.439 |
| S4-F1 | -0.168 | +0.068 | -0.050 |
| S4-F3 | +0.297 | +0.474 | +0.386 |
| S5-F1 | +0.506 | +0.270 | +0.388 |
| S5-F3 | +0.405 | +0.496 | +0.451 |
F2 omitted from table (near-constant HR). Mean smoothed r across all 30 pairs: +0.220. Range: -0.940 to +0.897.
Seven of 30 pairs exceed 95% of session-swap surrogates. K-MPAI vs Mean Smoothed r: rs = +0.000, p = 1.000. Trend-level coherence is not associated with anxiety level, suggesting shared environmental response rather than anxiety-driven coupling.
Smoothed Correlation by Condition: Prepared vs. Assigned Song
Part 10: Wavelet Coherence
Wavelet coherence decomposes both HR signals into time-frequency space using a continuous Morlet wavelet transform, measuring coherence (0 to 1) at each combination of time and period (timescale). This reveals where in the song and at what timescale the coupling is strongest.
Periods analyzed: 6 to 58 seconds (limited by song duration). The coherence heatmaps show coherence concentrated at longer timescales (30-58s periods), confirming that singer-panelist coupling is a slow, broad phenomenon.
| Pair | Song 1 Coh. | Song 2 Coh. | Mean Coh. |
|---|---|---|---|
| S1-F1 | 0.557 | 0.297 | 0.427 |
| S1-F3 | 0.469 | 0.312 | 0.391 |
| S2-F1 | 0.443 | 0.394 | 0.419 |
| S2-F3 | 0.605 | 0.399 | 0.502 |
| S3-F1 | 0.706 | 0.651 | 0.679 |
| S3-F3 | 0.554 | 0.642 | 0.598 |
| S4-F1 | 0.435 | 0.466 | 0.451 |
| S4-F3 | 0.487 | 0.454 | 0.471 |
| S5-F1 | 0.514 | 0.388 | 0.451 |
| S5-F3 | 0.511 | 0.412 | 0.462 |
Mean coherence across all 30 pairs: 0.489. Six of 30 pairs exceed 95% of surrogates. S3-F1 song_1 shows the highest coherence (0.706, z = +3.95). K-MPAI vs Wavelet Coherence: rs = -0.100, p = 0.873.
How the Three Synchrony Analyses Relate
| Analysis | Timescale | What It Detects | Key Finding |
|---|---|---|---|
| WCC (Part 8) | 5-10s fluctuations | Moment-to-moment coupling after detrending | Mostly null; few pairs exceed surrogates |
| Smoothed r (Part 9) | 15s+ trends | Broad HR trajectory co-movement | Strong coherence in many pairs (mean r = +0.22) |
| Wavelet (Part 10) | 6-58s decomposed | Multi-scale coherence, localized in time | Moderate coherence concentrated at 30s+ periods |
The three analyses converge: singer and panelist HR share broad arcs over the course of a song, but this shared trajectory is not driven by anxiety level and does not extend to moment-to-moment fluctuation coupling.
Part 11: Full-Session Coherence (Pre-Singing Through Post-Session)
Parts 8-10 analyzed synchrony during the two song phases only. But the aligned HR data for each session spans five phases: pre-singing, song 1, inter-song gap, song 2, and post-session. The pre-singing phase captures anticipatory activation (singers at 126-143 bpm before singing a note), the inter-song gap captures panelist anticipatory ramps (Panelist 3 ramping +17.7 bpm before Singer 4’s second song), and the post-session phase captures any immediate recovery. These are some of the most interesting periods for co-regulation evidence, and they fall outside the song windows analyzed in Parts 8-10.
Why extend the analysis? The song-phase synchrony results (Parts 9-10) showed trend-level co-movement during active singing (mean smoothed r = +0.22, mean wavelet coherence = 0.489). But they couldn’t answer a critical question: is this co-movement part of a broader shared physiological arc across the entire audition, or is it specific to the performance period? If singer and panelist HR co-move before, during, and after singing, that strengthens the case for a shared autonomic response to the audition environment. If the co-movement is song-specific, the non-performance phases tell a different story.
What we ran: Smoothed whole-series correlation and wavelet coherence on the full session time series (pre-singing through post-session) for each singer-panelist pair, with surrogate testing. WCC (Part 8) was not extended because the non-song phases are too short (pre-singing as brief as 17s, inter-song gap as brief as 18s) and WCC was already mostly null during the longer song phases.
Full-Session Durations (per subject)
| Session | Pre-Singing | Song 1 | Inter-Song Gap | Song 2 | Post-Session | Total |
|---|---|---|---|---|---|---|
| S1 | 17s | 207s | 25s | 59s | 10s | 318s |
| S2 | 41s | 85s | 40s | 57s | 15s | 238s |
| S3 | 39s | 147s | 20s | 56s | 13s | 276s |
| S4 | 53s | 165s | 32s | 55s | 13s | 319s |
| S5 | 84s | 165s | 18s | 59s | 12s | 340s |
Full-Session Smoothed Correlation
| Pair | Full-Session r | Song-Only Mean r | Delta |
|---|---|---|---|
| S1-F1 | +0.294 | -0.224 | +0.518 |
| S1-F3 | +0.450 | -0.202 | +0.651 |
| S2-F1 | -0.606 | -0.002 | -0.605 |
| S2-F3 | +0.198 | +0.771 | -0.573 |
| S3-F1 | -0.387 | +0.660 | -1.046 |
| S3-F3 | +0.185 | +0.439 | -0.254 |
| S4-F1 | -0.228 | -0.050 | -0.179 |
| S4-F3 | +0.082 | +0.386 | -0.304 |
| S5-F1 | +0.286 | +0.388 | -0.102 |
| S5-F3 | +0.487 | +0.451 | +0.036 |
F2 omitted from table (near-constant HR within songs, though included in full-session analysis where slow drift produces higher variability). Mean full-session smoothed r across all 15 pairs: +0.002, vs. song-only +0.220. Mean shift: -0.218.
The positive co-movement observed during songs collapses to near zero when the full session is analyzed. Song-only and full-session r values are essentially uncorrelated (r_s = +0.15, p = 0.594). The trend coherence during active singing is not part of a broader shared physiological arc. The anticipatory, transitional, and recovery phases introduce divergent patterns (asymmetric anticipatory responses, different recovery trajectories) that wash out the during-performance signal.
Full-Session Wavelet Coherence
| Pair | Full-Session Coh. | Song-Only Mean Coh. | Delta |
|---|---|---|---|
| S1-F1 | 0.602 | 0.427 | +0.175 |
| S1-F3 | 0.521 | 0.391 | +0.131 |
| S2-F1 | 0.499 | 0.419 | +0.081 |
| S2-F3 | 0.566 | 0.502 | +0.064 |
| S3-F1 | 0.615 | 0.679 | -0.064 |
| S3-F3 | 0.508 | 0.598 | -0.090 |
| S4-F1 | 0.559 | 0.451 | +0.108 |
| S4-F3 | 0.513 | 0.471 | +0.042 |
| S5-F1 | 0.576 | 0.451 | +0.125 |
| S5-F3 | 0.563 | 0.462 | +0.102 |
Mean full-session wavelet coherence: 0.565 (vs. song-only 0.489, +0.075 increase). Three of 15 pairs exceed 95% of surrogates.
Wavelet coherence moves in the opposite direction from smoothed correlation: it increases at the full-session level. This is not contradictory. Wavelet coherence captures coupling at specific timescales and time points, including the transitions between phases (onset of singing, inter-song gap, end of performance). These shared time-locked events boost coherence at specific scales even when the overall linear trend correlation is near zero.
What This Tells Us
Linear trend co-movement is song-specific. The mean positive correlation during songs (+0.220) disappears at the full-session level (+0.002). When a singer’s HR ramps up in anticipation, the panelists’ HR may or may not do the same, and the patterns differ across sessions.
Scale-specific coherence extends beyond songs. Coherence at 30s+ timescales persists and strengthens at the full-session level. The phase transitions themselves create shared time-locked events that register as coherence.
The distinction matters for co-regulation claims. The full-session results suggest something more modest than full-arc co-regulation: singer and panelists share a common response to the audition structure (event-locked coherence) and show directional co-movement during active singing (trend coherence), but neither pattern extends cleanly across the full session. This is consistent with a shared environmental response rather than direct interpersonal entrainment.
Full-Session vs. Song-Only Comparison
Summary
| Singer | Pre-Singing HR | Song 1 HR | Song 2 HR | Reactivity (Song 1 - Baseline) | Activation Gap (Song 2) | K-MPAI | Notes |
|---|---|---|---|---|---|---|---|
| S1 (M) | 128 | 144 | 154 | +16 | +70 | 126 | Steady climber; flattest assigned-song trace |
| S2 (F) | 128 | 160 | 147 | +32 | +66 | 111 | Steepest ramp-up; HR drops for song 2 |
| S3 (F) | 126 | 150 | 152 | +24 | +70 | 144 | Moderate, steady; similar to S2 during assigned song |
| S4 (F) | 91 | 106 | 90 | +15 | +8 | 159 | Outlier – pre-existing relationship with F3; HR near panelist range |
| S5 (M) | 143 | 167 | 170 | +24 | +91 | 174 | Highest absolute HR; no recovery between songs |
All values are mean HR in bpm. Activation gap = singer mean HR minus mean of all three panelists during song 2.
Part 12: Limitations and Design Considerations
This section consolidates the study’s limitations. Per-analysis limitations are detailed within their respective parts.
- Data resolution: HR was exported from Polar Flow at 1 Hz (averaged BPM per second), not beat-to-beat RR intervals. Standard HRV metrics cannot be computed. The 1 Hz resolution also limits within-window variability for WCC (Part 8), contributing to Panelist 2’s 95% exclusion rate.
- Sample size: With 5 singers, Spearman correlations cannot reach p < 0.05 unless the rank ordering is perfectly monotonic. All correlations are descriptive.
- Pre-existing relationships: Panelist 3 is Singer 4’s studio voice teacher. Panelist 1 had prior classroom teaching experience with all five participants. These relational histories are confounds that cannot be fully separated from individual autonomic differences.
- No medication screening: The study did not screen for beta blockers, SSRIs, or benzodiazepines, which could attenuate HR responses.
- Order confound: Song 2 (assigned) always follows song 1 (prepared), so any assigned-piece effect is inseparable from a second-performance effect.
- Gendered repertoire: Female and male singers sang different assigned pieces, limiting cross-gender comparison.
- No pianist: All performances used backing tracks (unintentional but serves as a control).
- Multiple comparisons: Over 50 correlations computed across Parts 4-11. Even the one p = 0.037 result (K-MPAI x HR Range) would not survive Bonferroni correction.